Lightpanda CVE-2026-52842

Exploitation requires the Lightpanda browser (version < 0.3.1) to fetch an attacker-crafted URL whose path contains an '@' followed by the target hostname, e.g. http://attacker.com/@victim.com/ - the '@' in the path segment is what causes the host to be misparsed. The limiting factor is user/agent interaction (UI:R): the victim automation must be induced to navigate to the malicious URL rather than the flaw firing unprompted, and the payload only yields impact against origins the browser holds sensitive context for (cookies, storage, authenticated sessions). No authentication is required (PR:N) and it works against default configurations, so any Lightpanda instance that processes untrusted or attacker-influenced URLs is exposed.

The signals are largely consistent toward high real-world severity for affected deployments. CVSS 3.1 is 9.3 with vector AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N - network reachable, low complexity, unauthenticated (PR:N), but requiring user/agent interaction (UI:R) to visit the crafted URL, with a scope change (S:C) reflecting that one origin's protections are subverted to affect another. There is no EPSS score in the provided data and no CISA KEV listing, and no public exploit was identified, so exploitation appears theoretical/opportunistic rather than mass-automated at this time. The one caveat worth flagging: the impact is a same-origin-policy/authentication-context bypass (confidentiality and integrity of cross-origin data), which the S:C/C:H/I:H rating captures well, though A:N correctly notes no availability impact. Overall this is a genuine priority for anyone running Lightpanda against untrusted content, not an inflated-CVSS non-issue, because the trigger condition is a single crafted URL.

An attacker hosts a page at http://attacker.com/@victim.com/ and lures a Lightpanda-driven automation agent or scraper into fetching it (UI:R). Lightpanda serves the attacker's content but stamps it with the origin http://victim.com, so the attacker's script can read cookies, storage, or authenticated responses scoped to victim.com and exfiltrate them, effectively bypassing the Same-Origin Policy. No public POC was identified, but the fix's own test cases (e.g. http://evil.example.com/@victim.example.com/) demonstrate the exact trigger.